A TTFS-based energy and utilization efficient neuromorphic CNN accelerator.

Spiking neural networks (SNNs), which are a form of neuromorphic, brain-inspired AI, have the potential to be a power-efficient alternative to artificial neural networks (ANNs). Spikes that occur in SNN systems, also known as activations, tend to be extremely sparse, and low in number. This minimizes the number of data accesses typically needed for processing. In addition, SNN systems are typically designed to use addition operations which consume much less energy than the typical multiply and accumulate operations used in DNN systems. The vast majority of neuromorphic hardware designs support rate-based SNNs, where the information is encoded by spike rates. Generally, rate-based SNNs can be inefficient as a large number of spikes will be transmitted and processed during inference. One coding scheme that has the potential to improve efficiency is the time-to-first-spike (TTFS) coding, where the information isn't presented through the frequency of spikes, but instead through the relative spike arrival time. In TTFS-based SNNs, each neuron can only spike once during the entire inference process, and this results in high sparsity. The activation sparsity of TTFS-based SNNs is higher than rate-based SNNs, but TTFS-based SNNs have yet to achieve the same accuracy as rate-based SNNs. In this work, we propose two key improvements for TTFS-based SNN systems: (1) a novel optimization algorithm to improve the accuracy of TTFS-based SNNs and (2) a novel hardware accelerator for TTFS-based SNNs that uses a scalable and low-power design. Our work in TTFS coding and training improves the accuracy of TTFS-based SNNs to achieve state-of-the-art results on the MNIST and Fashion-MNIST datasets. Meanwhile, our work reduces the power consumption by at least 2.4×, 25.9×, and 38.4× over the state-of-the-art neuromorphic hardware on MNIST, Fashion-MNIST, and CIFAR10, respectively.

Rectified Linear Postsynaptic Potential Function for Backpropagation in Deep Spiking Neural Networks.

Spiking neural networks (SNNs) use spatiotemporal spike patterns to represent and transmit information, which are not only biologically realistic but also suitable for ultralow-power event-driven neuromorphic implementation. Just like other deep learning techniques, deep SNNs (DeepSNNs) benefit from the deep architecture. However, the training of DeepSNNs is not straightforward because the well-studied error backpropagation (BP) algorithm is not directly applicable. In this article, we first establish an understanding as to why error BP does not work well in DeepSNNs. We then propose a simple yet efficient rectified linear postsynaptic potential function (ReL-PSP) for spiking neurons and a spike-timing-dependent BP (STDBP) learning algorithm for DeepSNNs where the timing of individual spikes is used to convey information (temporal coding), and learning (BP) is performed based on spike timing in an event-driven manner. We show that DeepSNNs trained with the proposed single spike time-based learning algorithm can achieve the state-of-the-art classification accuracy. Furthermore, by utilizing the trained model parameters obtained from the proposed STDBP learning algorithm, we demonstrate ultralow-power inference operations on a recently proposed neuromorphic inference accelerator. The experimental results also show that the neuromorphic hardware consumes 0.751 mW of the total power consumption and achieves a low latency of 47.71 ms to classify an image from the Modified National Institute of Standards and Technology (MNIST) dataset. Overall, this work investigates the contribution of spike timing dynamics for information encoding, synaptic plasticity, and decision-making, providing a new perspective to the design of future DeepSNNs and neuromorphic hardware.

Chemical speciation and fate of tripolyphosphate after application to a calcareous soil.

Adsorption and precipitation reactions often dictate the availability of phosphorus in soil environments. Tripolyphosphate (TPP) is considered a form of slow release P fertilizer in P limited soils, however, investigations of the chemical fate of TPP in soils are limited. It has been proposed that TPP rapidly hydrolyzes in the soil solution before adsorbing or precipitating with soil surfaces, but in model systems, TPP also adsorbs rapidly onto mineral surfaces. To study the adsorption behavior of TPP in calcareous soils, a short-term (48 h) TPP spike was performed under laboratory conditions. To determine the fate of TPP under field conditions, two different liquid TPP amendments were applied to a P limited subsurface field site via an in-ground injection system. Phosphorus speciation was assessed using X-ray absorption spectroscopy, total and labile extractable P, and X-ray diffraction. Adsorption of TPP to soil mineral surfaces was rapid (< 48 h) and persisted without fully hydrolyzing to ortho-P. Linear combination fitting of XAS data indicated that the distribution of adsorbed P was highest (~ 30-40%) throughout the site after the first TPP amendment application (high water volume and low TPP concentrations). In contrast, lower water volumes with more concentrated TPP resulted in lower relative fractions of adsorbed P (15-25%), but a significant increase in total P concentrations (~ 3000 mg P kg soil) and adsorbed P (60%) directly adjacent to the injection system. This demonstrates that TPP application increases the adsorbed P fraction of calcareous soils through rapid adsorption reactions with soil mineral surfaces.

Citrate Addition Increased Phosphorus Bioavailability and Enhanced Gasoline Bioremediation.

Phosphorus (P) bioavailability often limits gasoline biodegradation in calcareous cold-region soils. One possible method to increase P bioavailability in such soils is the addition of citrate. Citrate addition at the field scale may increase hydrocarbon degradation by: (i) enhancing inorganic and organic P dissolution and desorption, (ii) increasing hydrocarbon bioavailability, and/or (iii) stimulating microbial activity. Alternatively, citrate addition may inhibit activity due to competitive effects on carbon metabolism. Using a field-scale in situ biostimulation study, we evaluated if citrate could stimulate gasoline degradation and what the dominant mechanism of this stimulation will be. Two large bore injectors were constructed at a site contaminated with gasoline, and a biostimulation solution of 11 mM MgSO, 1 mM HPO, and 0.08 mM HNO at pH 6.5 in municipal potable water was injected at ∼5000 L d for about 4 mo. Following this, 10 mM citric acid was incorporated into the existing biostimulation solution and the site continued to be stimulated for 8 mo. After citrate addition, the bioavailable P fraction in groundwater and soil increased. Iron(II) groundwater concentrations increased and corresponded to decreases in benzene, toluene, ethylbenzene, xylenes (BTEX) in groundwater, as well as a decrease in F1 in the soil saturated zone. Overall, citrate addition increased P bioavailability and may stimulate anaerobic microbial activity, resulting in accelerated anaerobic gasoline bioremediation in cold-region calcareous soils.

Total Phosphate Influences the Rate of Hydrocarbon Degradation but Phosphate Mineralogy Shapes Microbial Community Composition in Cold-Region Calcareous Soils.

Managing phosphorus bioaccessibility is critical for the bioremediation of hydrocarbons in calcareous soils. This paper explores how soil mineralogy interacts with a novel biostimulatory solution to both control phosphorus bioavailability and influence bioremediation. Two large bore infiltrators (1 m diameter) were installed at a PHC contaminated site and continuously supplied with a solution containing nutrients and an electron acceptor. Soils from eight contaminated sites were prepared and pretreated, analyzed pretrial, spiked with diesel, placed into nylon bags into the infiltrators, and removed after 3 months. From XAS, we learned that three principal phosphate phases had formed: adsorbed phosphate, brushite, and newberyite. All measures of biodegradation in the samples (in situ degradation estimates, mineralization assays, culturable bacteria, catabolic genes) varied depending upon the soil's phosphate speciation. Notably, adsorbed phosphate increased anaerobic phenanthrene degradation and bzdN catabolic gene prevalence. The dominant mineralogical constraints on community composition were the relative amounts of adsorbed phosphate, brushite, and newberyite. Overall, this study finds that total phosphate influences microbial community phenotypes whereas relative percentages of phosphate minerals influences microbial community genotype composition.

Benzene removal by a novel modification of enhanced anaerobic biostimulation.

A novel modification of enhanced anaerobic bioremediation techniques was developed by using non-activated persulfate to accelerate the organic phosphorus breakdown and then stimulate benzene biodegradation by nitrate and sulfate reduction. Benzene concentrations in groundwater where nitrate, triethyl phosphate and persulfate were successfully injected were reduced at removal efficiencies greater than 77% to the levels below the applicable guideline. Soil benzene was removed effectively by the modification of the enhanced anaerobic bioremediation with removal efficiencies ranging between 75.9% and 92.8%. Geochemical analytical results indicated that persulfate effectively breaks down triethyl phosphate into orthophosphate, thereby promoting nitrate and sulfate utilization. Microbial analyses (quantitative polymerase chain reaction, denaturing gradient gel electrophoresis and 16S ribosomal RNA) demonstrated that benzene was primarily biodegraded by nitrate reduction while sulfate reduction played an important role in benzene removal at some portions of the study site. Enrichment in the heavier carbon isotope ¹³C of residual benzene with the increased removal efficiency provided direct evidence for benzene biodegradation. Nitrogen, sulfur and oxygen isotope analyses indicated that both nitrate reduction and sulfate reduction were occurring as bioremediation mechanisms.